NowComment
2-Pane Combined
Comments:
Full Summaries Sorted

第十二组 13 principles for managing essential

13 Principles for Managing Essential

Processing in Multimedia

Learning: Segmenting, Pre-

training, and Modality Principles

Richard E. Mayer

University of California, Santa Barbara

Celeste Pilegard

University of California, Santa Barbara

Abstract

 When a concise multimedia lesson containing complicated material is presented at a fast rate, the result can be a form of cognitive overload called essential overload . Essential overload occurs when the amount of essential cognitive processing (similar to intrinsic cognitive load) required to understand the multimedia instructional message exceeds the learner’s cognitive capacity. Three multimedia design methods intended to minimize essential overload are the segmenting, pre-training, and modality principles. The segmenting principle is that people learn more deeply when a multimedia message is presented in learner-paced segments rather than as a continuous unit. This principle was supported in 10 out of 10 experimental tests, yielding a median effect size of 0.79. The pre-training principle is that people learn more deeply from a multimedia message when they know the names and characteristics of the main concepts. This principle was supported in 13 out of 16 experimental tests, yielding a median effect size of 0.75. The modality principle is that people learn more deeply from a multimedia message when he words are spoken rather than printed. This principle was supported in 53 out of 61 experimental tests, yielding a median effect size of 0.76.

What Are the Segmenting, Pre-training, and Modality

Principles?

Definitions

Consider a multimedia learning situation in which too much essential information is presented at too fast a rate for you to adequately process. For example, when a narrated animation contains a large amount of complicated material that is not familiar to you and comes to you at a fast pace determined by the computer, the result can be a form of cognitive overload called essential overload . Essential overload occurs when the amount of essential cognitive processing required by the multimedia instructional message exceeds the learners cognitive capacity. Essential processing (similar to Swellers intrinsic cognitive load ; Sweller, Ayres, & Kalyuga , 2011 ; see also Chapter 2 ) refers to cognitive processing such as selecting relevant words and images and organizing them as presented (as described in Chapter 3 ) required to represent the essential material.

1Cognitive capacity refers to the total amount of processing that can be supported by both the auditory and visual channels of the learners working memory at any one time . Essential material refers to the words and pictures needed to achieve the instructional objective, such as all words and graphic elements needed for understanding how a lightning storm develops.

To address the problem of essential overload, we cannot reduce the amount of essential material (analogous to reducing extraneous processing in the preceding chapter), because helping students learn the essential material is the instructional goal. Thus, what we need are some techniques for managing essential processing that is, for helping students process the essential material without experiencing essential overload. The goal of this chapter is to examine the research evidence concerning three principles of multimedia design aimed at minimizing the effects of essential overload the segmenting principle, the pre-training principle, and the modality principle. The segmenting principle is that people learn more deeply when a multimedia message is presented in learner-paced segments rather than as a continuous unit. The pre-training principle is that people learn more deeply from a multimedia message when they know the names and characteristics of the main concepts. The modality principle 2 is that people learn more deeply from a multimedia message when the words are spoken rather than printed.

1Throughout this chapter, I use the term essential cognitive processing to refer to largely the same concept as Swellers intrinsic cognitive load ( Sweller, Ayres, & Kalyua , 2011 ; see also Chapter 2 )

2 Sweller ( Sweller, Ayres, & Kalyua , 2011 ; see also Chapter 9 ) considers the modality effect to be related to the split-attention effect in that both require two or more sources of information that cannot be understood in isolation and so must be mentally or physically integrated. Split-attention effects include spatial contiguity and temporal contiguity, which are examined in the preceding chapter on extraneous overload. We agree that modality effects and split-attention effects derive from the same problem; in particular, modality effects can occur when the learner must focus on two kinds of visual information presented at the same time printed words and graphics which may overload the learners visual channel. We have chosen to include modality in this chapter because it corresponds to a form of essential overload (according to the taxonomy in Figure 3.3 of Chapter 3 ) having too much material presented at one time to the visual system. Split-attention effects are examined in the preceding chapter because we attribute them to confusing layout (a form of extraneous overload). Thus, there appears to be no major disagreement between Swellers analysis of the modality principle (as summarized in Chapter 9 ) and ours.

What Is the Theoretical Rationale for the segmenting,

Pre-training, and Modality Principles

 A major challenge for instructional designers is to create instructional messages that are sensitive to the characteristics of the human information processing system. In particular, the cognitive theory of multimedia learning summarized in Figure 3.2 of Chapter 3 shows that much of the cognitive processing for meaningful learning occurs within working memory. According to the cognitive theory of multimedia learning, the visual/pictorial and auditory/verbal channels in working memory are extremely limited, so that only a few items can be held or manipulated in each channel at any one time. When an instructional message such as a narrated animation presents a lot of unfamiliar essential material to the learner at a rapid rate, the cognitive capacity of the information processing system can become overloaded a situation we call essential overload . Carrying out cognitive processing takes time, but a fast-paced presentation that requires a lot of mental model building may not allow enough time. As a result, the learner may not be able to engage in all of the cognitive processing needed for making sense of the presented material, so full understanding may not be achieved ( Mayer, 2009 ; Mayer & Moreno , 2003 ).

Table 13.1 presents two overload scenarios, each involving a form of essential overload (Mayer & Moreno, 2003 ). In the first scenario (called type 1 essential overload ), both channels are overloaded by essential processing demands (or intrinsic cognitive load). This overload scenario can occur when a narrated animation concerning a complex topic is presented at a fast pace. Two load-reducing methods are segmenting (breaking the lesson into manageable learner-controlled segments) and pre-training (providing the names and characteristics of key components before the lesson), both of  which are explored in this chapter. The theoretical rationale for segmenting is that it slows the pace of presentation to a level that enables learners to carry out essential processing. The theoretical rationale for pre-training is that it equips learners with prior knowledge that they can use to process the subsequent narrated animation with less cognitive effort. Thus, segmenting gives them the time they need to carry out essential processing, whereas pre-training reduces the amount of essential processing that is required.

In the second overload scenario (called type 2 essential overload ), the

IMG_256
visual channel is overloaded by essential processing demands. This overload scenario can occur when a lesson with animation and concurrent on-screen text (or with static diagrams and printed text) is presented at a fast pace. A load-reducing method is to off-load the verbal processing from the visual channel to the auditory channel by presenting the words as narration rather than as on-screen text. This approach is also examined in Chapter 9 on the modality principle .

Examples of the Segmenting, Pre-training,

and Modality Principles

How can we design multimedia instructional messages such as narrated animations so that they do not create essential overload?

In other words, what can we do to a multimedia instructional message to manage the amount of essential cognitive processing that is required to take place at any one time? In this chapter, we explore three techniques for reducing essential overload segmenting techniques, pre-training techniques, and modality techniques.

Segmenting Techniques

Lets begin with a narrated animation such as a 140-second narrated animation explaining the steps in the formation of lightning. Frames from the narrated animation are shown in Figure 3.1 of Chapter 3 . The explanation is complex, consisting of more than a dozen steps and including more than a dozen interacting elements. Students must focus on the key words and images (such as moist cool air coming into contact with a warm surface), must note how a state change in one element causes another change (such as noting that the air rises when it becomes heated), and must relate the events to prior knowledge (such as knowing that heat causes a gas to expand and to thereby become relatively lighter). Many students particularly those with low levels of prior knowledge may have difficulty keeping up with the pace of the presentation so they are not able to engage in all of the needed processing.

IMG_256

  One solution to this essential overload problem is to allow the learner to control the pace of presentation. For example, we can break the 140-second narrated animation on lightning into 16 segments, each lasting about 10 seconds and consisting of a sentence or two. Table 13.2 shows how we can break the script of the lightning lesson into 16 segments. Further, as indicated in Figure 13.1 , we can put a Continuebutton in the lower right corner of the screen that appears at the end of each animation segment. The learner can use the mouse to click on Continuewhenever the learner is ready to go on to the next segment. When the learner has digested one segment that is, when the learner has engaged in the cognitive processes shown in Figure 3.2 of Chapter 3 the learner can move on to the next segment. In this way, the learner has some modest control over the pace of presentation of the narrated animation, and thus a way of avoiding the problem of not having enough time to carry out the required cognitive processing. It is important that each verbal segment be short involving 810 seconds of speech concerning one main event so the learner can easily hold the words in working memory while viewing the corresponding animation segment.

IMG_256

As a second example, consider a narrated animation that explains how an electric motor works. The motor consists of many unfamiliar parts – such as a wire loop, commutator, magnet, battery, and wires – and many causal links. For an inexperienced learner, the pace of presentation may not allow for the complete cognitive processing needed to build a meaningful mental model. To alleviate this problem, we can put a list of questions in the upper right corner of the screen – for example, “What happens when the motor is in the start position?” “What happens when the motor has rotated a quarter turn?” “What happens when the motor has rotated a half turn?” “What happens when the motor has rotated three quarters of a turn?” “What happens when the motor has rotated a full turn?” The learner can click on any of the elements in the electric motor – IMG_256
such as the magnet or wire loop or commutator or wires or battery – and then click on any of the questions. A short segment of narrated animation or annotated animation addressing this question then appears, such as shown in Figure 13.2 . In this way, the learner sees exactly the same narrated animation as in the continuous presentation but can control the pace and order of the presentation .

Pre-training Techniques

Another solution to the essential overload problem is to equip the learner with knowledge that will make it easier to process a narrated animation. For example, suppose we present a narrated animation explaining how a car’s braking system works. The script for the lesson and some selected frames are shown in Figure 13.3 . Mayer, Mathias, and Wetzell ( 2002 ) proposed a two-stage learning process in which learners first build component models for each major part in the system and then build a causal model. Building component models consists of learning the name and behavior of each component, such as learning that the piston in the master cylinder can move forward or backward, the brake fluid in the tube can be compressed or not compressed, and so on. Building a causal model consists of learning the causal chain, such as stepping on the car’s brake pedal causes a piston to move forward in the master cylinder, which in turn causes brake fluid in the tube to compress, and so on.

The pace of presentation may be so fast that by the time learners are able to build component models, there is no time left to build a causal model. To overcome this overload problem, we can provide pre-training to the learners concerning the names and characteristics of each component. For example, Figure 13.4 shows frames from a pre-training episode in which learners can click on any part of a diagram of the braking system such as the piston in the master cylinder and then be given the name for that part and shown the states the part can be in. After the learner has clicked on each part, the learner can be shown the narrated animation explaining how a cars braking system works. However, because the learner already knows the name and characteristics of each part, the learner can engage in cognitive processes for building a causal model of the system, leading to better understanding. In this way, the pre-training provides prior knowledge that reduces the amount of processing needed to understand the narrated animation. If learners

already possess prior knowledge, pre-training is not needed.

IMG_256

Modality Techniques

Finally, suppose you are given an animation explaining lightning formation, along with captions (one or two sentences in length) at the bottom of the screen that describe the events rendered in the animation (such as shown in the top half of Figure 13.5 ). In this situation, the words in the multimedia message are presented as on-screen text. This situation may overload the visual channel because the learner must look at both the animation and the on-screen text at the same time. In order to off-load some of the visual processing, we can present the words as concurrent narration (as shown in the bottom half of Figure 13.5 ). In this way, learners can watch the lightning animation with their eyes and listen to the verbal explanation of lightning formation with their ears.

IMG_256


What Do We Know about the Segmenting, Pre-training, and Modality Principles?

The previous edition of this handbook reported a small but consistent research base supporting the segmenting, pre-training, and modality principles. The current chapter shows that the research base has more


than doubled since the previous edition, but the support for the segmenting, pre-training, and modality principles remains near the high range. The major new contributions concern the search for theory-based boundary conditions under which the principles are most likely to apply or not apply.

In this review, we consider articles published in archival journals or chapters in which (1) the task is an instructional lesson involving words and pictures; (2) the independent variable involves whether or not the presentation was segmented (i.e., continuous vs. segmented), whether or not pre-training was provided on elements in the presentation (i.e., pre-training vs. no pretraining), or whether or not the words are in spoken or printed form (i.e., spoken text vs. printed text); (3) the dependent measure is performance on a transfer test; and (4) reported statistics include the mean and standard deviation of each group. We show effect sizes separately for low-knowledge and high-knowledge learners when prior knowledge was explicitly manipulated, recognizing that strong positive effects are not expected for high-knowledge learners. We computed effect sizes for each comparison by subtracting the mean problem-solving transfer score of the control group from the mean problem-solving transfer score of the experimental group and dividing by the pooled standard deviation (Ellis, 2010 ). Following Cohen (1988), we consider an effect size of d = 0.20 to be small, d = 0.50 to be medium, and d = 0.80 to be large; and following Hattie (2009) , we consider any effect size greater than d = 0.40 to be educationally important .

Research on the Segmenting Principle

Do students learn more deeply when an information-rich multimedia lesson is presented in learner-paced segments rather than as a continuous unit? Table 13.3 summarizes 10 comparisons between a group that received a multimedia presentation broken into segments with the pace controlled by the learner (segmented group) and a group that received the same multimedia presentation as a single continuous presentation (continuous group). The left side of the table lists the experiment that is the source of the data, the middle portion of the table lists the content of the lesson, and the right side of the table lists the effect size. Overall, in 10 of 10 comparisons, there was a positive effect size based on transfer test score, with a median effect size of d = 0.79.   

In the first study listed in Table 13.3, Mayer and Chandler (2001)  compared the learning outcomes of students who viewed a 140-second narrated animation on lightning formation as a continuous presentation (continuous group) with the learning outcomes of students who viewed the same presentation in 16 segments (segmented group). Each segment lasted about 10 seconds and contained about a sentence of narration (as shown in Table 13.2 ); after the narrated animation segment was complete, a “Continue” button appeared on the screen (as shown in Figure 13.1) . When the learner clicked on the button, the next segment appeared. This procedure was repeated so the learner saw the continuous presentation twice or the segmented presentation twice. Line 1 of Table 13.3 shows that the segmented group performed better than the continuous group on a problem-solving transfer test, yielding a large effect size.

In a second set of studies, Mayer, Dow, and Mayer (2003) compared the learning outcomes of students who learned about electric motors from a simulation game in which they interacted with an on-screen agent named Dr. Phyz. In the continuous version, when the student clicked on the


electric motor, Dr. Phyz narrated a continuous animation showing how the electric motor works. In the segmented version, a list of questions appeared corresponding to segments of the narrated animation. When the student clicked on a question, Dr. Phyz narrated an animation concerning a segment of the presentation. When the segment was completed, the student could click on another question to see another segment of the narrated animation. As shown in lines 2 and 3 of Table 13.3 , the segmented group performed much better on a transfer test than the continuous group, yielding large effect sizes in both cases. 

Lines 4 and 5 summarize two experiments by Moreno (2007)  in which prospective teachers received a video lesson (Exp. 1) or animated lesson (Exp. 2) on teaching skills. Either the participants had to click to continue the lesson through seven segments (segmented group), or they saw the lesson straight through (continuous group). Across both experiments the segmenting group outperformed the continuous group on measures of transfer. 

In a study by Hasler, Kersten, and Sweller (2007) , elementary school students watched a narrated animation on the causes of day and night. The students viewed the lesson all at once (continuous group) or in learnerpaced predeined segments (segmented group). The segmented group outperformed the continuous group on low element interactivity questions (d=0.31), high element interactivity questions (d=1.20), and overall (d=0.81, as shown in line 6).

In a study by Lusk et al. (2009) , college students with low and high working memory capacity (WMC) viewed a segmented or continuous multimedia lesson on historical inquiry. Students with low-WMC in the segmented group outperformed the low-WMC students in the continuous group on the application of historical inquiry methods to a new source, as shown in line 7. Segmentation did not signiicantly affect performance for students with high WMC, and segmentation eliminated performance differences between low- and high-WMC students.

Line 8 summarizes a study by Boucheix and Schneider (2009) in which college students learned the functioning of a pulley system from animations that were either continuous or segmented in controllable microsteps. The segmented group outperformed the continuous group on a functional mental model test (d= 0.31). Boucheix and Guignard (2005) reported similar results for a study involving gears, but they did not provide enough information to allow for the computation of effect size. 

Line 9 summarizes a study by Stiller, Freitag, Zinnbauer, and Freitag (2009) in which college students were shown either a segmented or a continuous multimedia presentation on the structure of the human eye. The segmented group outperformed the continuous group on a transfer test when the words were printed (d= 0.30) but only slightly when the words were spoken (d= 0.06), so line 9 shows the combined effect size.                Finally, as summarized in the last line of Table 13.3 , Hassanabadi, Robatjazi, and Savoji (2011) gave middle school students a multimedia lesson on lightning formation that was either segmented or continuous. The segmented group had higher transfer scores (d= 0.24) for narrated lessons but only slightly higher transfer scores for lessons with on-screen text (d=0.09), so line 10 shows the combined effect size. 

Overall, Table 13.3 shows that there is consistent evidence for the segmenting principle: people learn more deeply when a multimedia message is presented in learner-paced segments rather than as a continuous unit. The median effect size for transfer test performance is near the large range, indicating that a modest design change that adds no new information can have a large effect on student understanding. 

Can the segmenting principle be extended beyond breaking a continuous presentation into learner-paced segments? First, there is some evidence that transfer test performance can be improved when a continuous multimedia lesson is broken into segments with short pauses (e.g., of 2 or 5 seconds) between them (Khacharem, Spanjers, Soudji, Kalyuga, & Pipoll , 2013 ; Singh, Marcus, & Ayres, 2012) . Further research is needed to disentangle the contributions of breaking a continuous presentation into bite-size segments and allowing learners to control the initiation of the next segment (also see Chapter 21 on learner control). In addition, research is needed to determine whether students learn better when they can determine the size of segments (such as by using a pause key) or when the instructor creates predetermined segments (e.g., as examined by   Boucheix & Schneider, 2009 ; Hasler, Kersten, & Sweller, 2007 ).

Second, there is emerging evidence that the segmenting principle may apply to the presentation of worked examples in mathematics, with greater transfer test performance when students see problems broken into meaningful steps rather than as a single formula ( Ayres , 2 006 ; Gerjets, Scheiter, & Catrambone , 2006 ). However, Spanjers and colleagues obtained smaller segmenting effects on transfer scores for animated or static worked examples of probability problems (Spanjers, van Gog, & van Merriënboer, 2012; Spanjers, van Gog, Wouters, & van Merriënboer,    2012; Spanjers, Wouters, van Gog, & van Merriënboer, 2011) . Further research is needed to address effective segmenting techniques for worked examples (also see Chapter 16 on worked examples).

Third, breaking complex data graphs into parts has been shown to improve transfer test performance both in the geosciences (Mautone & Mayer, 2007) and in chemistry (Lee, Plass, & Homer , 2006). Further research is needed to develop research-based design principles speciic to data graphs to complement expert advice (Wong, 2010 ). 

Some potential boundary conditions are that segmenting may have stronger effects for learners with low rather than high working memory capacity (Lusk et al., 2009) and for low-achieving rather than high-achieving learners (Ayres, 2006).   

Research on the Pre-training Principle

Do students understand a multimedia lesson better if they are provided with pre-training concerning the names and characteristics of the major elements in the lesson? Table 13.4 lists 16 tests of the pre-training principle. In the irst set of four studies (Pollock, Chandler, & Sweller, 2002) , apprentices took a course in electrical engineering that included a two-phase multimedia lesson on conducting safety tests for electrical appliances. For some learners (pretraining group), the irst phase focused on how each component worked, and the second phase focused on how all the components worked together within the electrical system. For other learners (no-pre-training group), both phases focused on how all the components worked together within the electrical system. On a subsequent problem-solving transfer test concerning how the elements worked together within the electrical system, learners with low levels of prior knowledge in the pre-training group performed better than students with low prior knowledge in the no-pre-training group, yielding large effect sizes in both cases, as shown in lines 1 and 3 of Table 13.4. In contrast, no advantage of pre-training was found for learners with high levels of prior knowledge, as shown in lines 2 and 4, indicating that high-experience learners were less likely to encounter essential overload. 

In the second set of three studies (Mayer, Mathias, & Wetzell, 2002) , students received a narrated animation explaining the workings of a car’s braking system or a bicycle tire pump and then took problem-solving transfer tests. Before the lesson on brakes, some students (pre-training group) received pre-training in which they learned the name and possible states of each component in the brake system – for example, the piston in the master cylinder could be forward or back, the luid in the brake tube could be compressed or not compressed, and so on. Before the tire pump lesson, some students (pretraining group) received pre-training with a clear plastic model in which they were asked to pull up and push down on the handle several times. On a subsequent test of problem-solving transfer, students in the pre-training group performed better than students in the


no-pre-training group across all three experiments, yielding large effect sizes, as shown in lines 5–7 of Table 13.4 . Next, in a set of two studies (Mayer, Mautone, & Prothero, 2002) , students learned about geology in a simulation game called the Proile Game. The goal of the game was to determine which geological feature was on a certain portion of the earth’s surface, represented as a window on the computer screen. Students could use a mouse to draw lines and were shown the depth or height at each point along the line. Some students (pre-training group) were shown illustrations of the major geological features – such as a ridge or a trench – before the lesson, whereas others (no-pre-training group) were not. As can be seen in lines 8 and 9, the pre-training group performed better on a subsequent test of problem-solving transfer than did the no-pretraining group, yielding medium to large effect sizes.

In a study by Clarke, Ayres, and Sweller (2005) , high school students learned about graphic representations of linear functions through a spreadsheet application. Students were divided by high and low spreadsheet skills and given instructions with either pre-training in spreadsheet skills or spreadsheet instruction presented concurrently with mathematics information. Pretraining resulted in higher performance for students with low spreadsheet skills (d= 1.84, as shown in line 10) but not for students with high spreadsheet skills (d= −0.38 as shown in line 11).

In the next series of four experiments, conducted by Kester and colleagues, students learned to solve practice problems in a computer simulation with supporting information provided either before the lesson (similar to pretraining) or during the lesson (similar to no pre-training) and then took a problem-solving transfer test. As can be seen in lines 12–15, pre-training was helpful in a study involving electrical circuits problems (Kester, Kirschner, & van Merriënboer, 2006) but not for other studies involving statistics problems (Kester, Kirschner, & van Merriënboer , 2004a ), electrical circuit problems (Kester, Kirschner, & van Merriënboer, 2004a) , or neural network problems (Kester, Lehnen, van Gerven, & Kirschner , 2006 ). 

Finally, the last line of T able 13.4 summarizes a study by Eitel, Scheiter, and Schüler (2013) in which college students learned about the structure and functioning of a pulley system by reading an instructional text. When a selfpaced diagram of the pulley system was presented before the lesson (as a form of pre-training), performance was improved on a comprehension test. McCrudden, Magliono, and Schraw (2011) also reported improvements on a recall test when a pre-training diagram was presented before a lesson on how kidney stones form during space travel, but this study is not included in the table because no transfer test was given. 

Overall, Table 13.4 shows that most studies produced effect sizes favoring the pre-training group, with a median effect size of d = 0.75. These indings are consistent with the pre-training principle: people learn more deeply from a multimedia message when they know the names and characteristics of the main concepts. These indings are consistent with earlier research on advance organizers (e.g., Mayer, 1983) showing that presenting brief, supporting information before a lesson can greatly increase transfer test performance. 

However, an important boundary condition suggested in Table 13.4 is that the pre-training principle may not apply to learners with high prior knowledge, perhaps because they are less likely to experience essential overload. The studies by Kester and colleagues ( Kester, Kirschner, & van Merriënboer, 2004a , 2004b , 2006 ; Kester, Lehnen, van Gerven, & Kirschner , 2006) suggest that more work may be needed to pinpoint how best to provide supporting information when students learn by solving problems with an interactive simulation.   

Research on the Modality Principle

Table 13.5summarizes more than 60 published experimental comparisons between the transfer test performance of a group that received graphics with printed text versus a group that received graphics with spoken text. As can be seen, the modality principle has been the focus of dozens of studies and, in fact, it is the most studied of all the multimedia learning principles. This section breaks modality research into three sections – foundational research published between 1995 and 2003 (as cited in the previous edition of this handbook), subsequent research testing boundary conditions, including subsequent positive research (i.e., which produced effect sizes of  d = 0.20 or greater), and subsequent negative research (i.e., which produced negative effect sizes or positive ones of less than d= 0.20). The i rst 33 lines of Table 13.5  summarize foundational i ndings, all of which show better transfer performance for multimedia lessons consisting of graphics with spoken text rather than graphics with printed text. As Low and Sweller describe in C hapter 9 , research on modality effects has a long history in cognitive psychology, but research on the modality principle in instructional contexts began in 1995 with a classic set of paper-based studies by Mousavi, Low, and Sweller(1995 ). Across five experiments, listed in lines 1–5, students performed much better on transfer tests after learning from paper-based lessons on how to solve geometry problems that were explained by a recorded voice rather than by text printed on the paper. Similar results were reported for paper-based lessons on electrical circuits in lines 6–8 (Tindall-Ford, Chandler, & Sweller, 1997) and graph reading on line 33 (Leahy, Chandler, & Sweller, 2003 ). T he remaining foundational studies in T able 13.5 involve computer-based lessons in which the graphics were presented as diagrams (Jeung, Chandler, & Sweller,1997 ;Kalyuga, Chandler, & Sweller , 1999 ,2000 ), as animation (Craig, Gholson, & Driscoll,2002;Mayer, Dow, & Mayer,2003 ;Mayer & Moreno,1998 ; Moreno & Mayer, 2002 , Exps. 1and 1b; Moreno et al,2001), or via a head-mounted display as virtual reality (Moreno & Mayer, 1999 2002 ;O’Neil et al,2000 ). Modality effects favoring graphics with narration over graphics with printed text were found across all foundational studies, including lessons on solving mathematics problems as summarized in lines 9–11 (Jeung, Chandler, & Sweller, 1 997) and in lines 24–26 (Atkinson ,2000); lightning formation in lines 12, 15, and 16 (Mayer & Moreno,1998;Moreno & Mayer,1999 ) and line 23 ( Craig, Gholson, & Driscoll ,2002); an environmental science game in lines 19, 20, 21, 22, 27, 28, 29, 30, and 31 ( M oreno et al., 2 001;     M oreno & Mayer, 2002) ; electrical engineering in lines 14 and 17 (Kalyuga, Chandler, & Sweller,1999,2000); how a car’s braking system works in line 13 (Mayer & Moreno, 1998) ; an aircraft simulation in line 18 (O’Neil et al,2 000 ); and how an electric motor works in line 32 (Mayer, Dow, & Mayer,2003) . Of the 33 foundational studies, all yielded positive modality effects, with a median effect size of d= 0.88.

Principles for Managing Essential Processing

IMG_256

IMG_256

IMG_256

IMG_256

In the ensuing years, research began to focus on boundary conditions for the modality principle, foreshadowed by Kalyuga, Chandler, and Sweller’s ( 2000 )finding that the modality effect was strong for learners with low prior knowledge but not for those with high prior knowledge. For example,Schmidt-Weigand, Kohnert, and Glowalla ( 2010a)found a modality effect when the lightning lesson was system-paced at a slow or medium pace but not when it was presented at a fast pace or self-paced (lines 41–44).Leahy and Sweller( 2011 ) and Wong, Leahy, Marcus, and Sweller( 2012 ) found a modality effect favoring narration when the words in a lesson on temperature graphs were presented in short segments.but a reverse modality effect favoring printed words when words were presented in long segments (lines 53–56). Thus, some potentially important boundary conditions are that the modality principle might not apply when the learners are knowledgeable, the lesson is self-paced, and the verbal segments are long. Positive modality effects were found in a biology classroom by Harskamp, Mayer, Suhre, and Jansma ( 2007), shown in lines 35 and 36; in a multimedia music theory lesson by Owens and Sweller (2008) , showin in line 37; in a lesson on learning to solve probability problems by Wouters, Paas, and van Merrië  nboer (2009 ), shown in line 38; in a replication using the lightning lesson in German by Schmidt-Weigand, Kohnert, and Glowalla ( 2010b) , in lines 45 and 46; in a biology lesson by Park, Moreno, Seufert, and Brunken ( 2011 ), in line 47; and in a multimedia lesson on i sh locomotion by K ü hl, Scheiter, Gerjests, and Edelmann ( 2011 ), in lines 51 and 52. T he modality effect was not found in a variety of situations, some of which are consistent with cognitive theory, as described by Low and Sweller in Chapter 9.For example, as shown in line 34,Tabbers, Martens, and van Merri ë nboer( 2004 ) were the first to find a negative modality effect for a lesson on instructional design, which may be attributed to the use of a selfpaced lesson, which can wipe out the benei ts of spoken text. As another example, the strongest negative modality effect was obtained for a lesson on tornados consisting of eight static slides with captions versus with narration (Schüler, Scheiter, Rummer, & Gerjets, 2012 ; see line 58). A clue to the result is rel ected in the fact that the negative modality effect favoring printed text was found for learners with low working memory capacity but not for those with high working memory capacity. This suggests that the transient nature of long verbal segments may be problematic when the presented words exceed what can be held in working memory for a slide. In short, the i nding that the modality principle might not apply when the lesson is self-paced or the verbal segments are long is consistent with the predictions of the cognitive theory of multimedia learning. T he remaining negligible or negative effects may also be explainable in terms of boundary conditions that are consistent with cognitive theory. As summarized in lines 39 and 40,Witteman and Segers (2010)found a modest modality effect for the lightning lessons when students received an immediate test but not for a delayed test, signaling the need for more research with delayed tests. Mayrath, Nihalani, and Robinson(2 011)did not i nd support for the modality principle in a computer networking simulation, as shown in lines 48 and 49. Lindow et al. (2011) , as summarized in in line 50, and Cheon, Crooks, and Chung ( 2013) , in line 61, did not i nd a modality effect for the lightning lesson, although it was modii ed by, for example, the addition of pauses between frames. Negative modality effects were reported by Crooks, Cheon, Ian, Ari, and Flores (2012) , as summarized in line 57, with a lesson on human speech and by Schüler, Schieter, Rummer, and Gerjets (2012) , line 58, with a lesson on tornados. Only small modality effects were reported by Schü l er, Scheiter, and Gerjets (2013) , in lines 59 and 60, for a biology lesson. Overall, across a wide variety of learning situations, the preponderance of evidence shows that people tend to learn better from graphics and spoken text than from graphics and printed text, with a median effect size of d= 0.76. These indings support the modality principle: people learn more deeply from multimedia messages when the words are presented as spoken text rather than printed text. The pattern of results in this review is consistent with an extensive metaanalysis of the modality effect conducted by Ginns (2005 ), yielding an overall weighted mean of d= 0.72 based on 43 independent comparisons. Consistent with the cognitive theory of multimedia learning, Ginns noted that the modality effect was strong for complex material but not for simple material (in which the learner may not experience cognitive load); for system-paced presentations but not for self-paced presentations (in which the effects of split attention can be overcome by spending sufi cient time to read the text and view the graphic); and on transfer tests rather than retention tests (consistent with the focus of this review).  Low and Sweller’s more recent review of the modality principle in Chapter 9 is consistent with this review and offers theory-based explanations for situations in which a modality effect is not observed (although Sweller and colleagues may classify the modality principle as involving extraneous rather than essential processing).  Reinwein(2 012 ) found moderate evidence for the modality principle, particularly for system-paced rather than selfpaced lessons, dynamic rather than static graphics, and transfer rather than retention measures. The growing number of failures to i nd a modality effect in recent years represents an important opportunity to ine-tune the cognitive theory of multimedia learning by identifying boundary conditions for when the effect is and is not found. According to the cognitive theory of multimedia learning, we would expect the modality principle to apply when the material is complex rather than simple, the presentation is system-paced rather than self-paced, the graphics are dynamic rather than static, the learners have a low level of knowledge rather than a high level, the verbal segments are short rather than long, and the words are familiar rather than unfamiliar.

The modality principle should not be taken to mean that spoken words are better than printed words in all situations. Printed words might be helpful when the verbal material contains technical terms, is in the learner’s second language, or is presented in segments that are too large to be held in the learner’s working memory. Instead of asking whether or not the modality effect exists, a much more promising direction is to identify theory-based boundary conditions for when it does and does not apply.    

What Are the Implications of Research for Cognitive Theory?

The research results summarized in Tables 13.3 , 13.4 , and 13.5  provide support for the predictions of the cognitive theory of multimedia learning as summarized in Figure 3.2 of Chapter 3. When learners do not have enough capacity available to engage in active cognitive processing of the essential material, their learning outcomes suffer – as indicated by tests of problem-solving transfer. According to the cognitive theory of multimedia learning, three ways to handle an essential overload situation are to allow the learner to slow down the pace of presentation (i.e., segmenting principle), provide the learner with knowledge that reduces the need for cognitive processing of the presentation (i.e., pre-training principle), or off-load some of the visual information onto the auditory channel (i.e., modality principle). Thus, each of these principles has theoretical plausibility because it was derived from the cognitive theory of multimedia learning – particularly concerning the limited capacity for processing information in working memory. The empirical evidence in support of each of these principles provides empirical plausibility and support for the predictions of the cognitive theory of multimedia learning. T he emerging body of research on boundary conditions can be useful in i ne-tuning the cognitive theory of multimedia learning. Progress can be made to the extent that boundary conditions can be explained in terms of cognitive theory. For example, we do not expect segmenting to be useful when the material is familiar and uncomplicated for the learner, because learners possess enough cognitive capacity to process the lesson. Spanjers, van Gog, & van Merri ë nboer ( 2010 ) propose that adding pauses to a systempaced presentation could serve the same cognitive function as segmenting, so adding a “Continue” button would not be expected to help. Similarly, we do not expect the pre-training principle to apply to learners with high prior knowledge because they already possess the information presented in pretraining. Finally, we do not expect the modality principle to apply when the lesson is self-paced (so learners can compensate for overloading situations), the verbal material is lengthy (so the transitory nature of spoken text is problematic), or when learners have a high level of prior knowledge (so they can process the lesson under a variety of conditions).   

 What Are the Implications of Research for Instructional Design?

T he research reviewed in this chapter shows that instructional designers should be sensitive to working memory constraints when presenting a complex multimedia lesson. For example, if a concise narrated animation contains a lot of interacting concepts and is presented at a fast pace, the demand for cognitive processing can exceed the learner’s cognitive capacity. Even if extraneous material has been eliminated from the presentation, the remaining essential material may be presented at a rate that exceeds the learner’s capacity. The segmenting, pre-training, and modality principles are particularly relevant to the design of narrated animations, narrated videos, or narrated slideshows that contain a lot of interacting concepts presented at a fast pace. According to the segmenting principle, it would be useful to break a narrated animation into meaningful segments and to allow the learner to control the onset of each segment, such as by clicking on a “Continue” button. According to the pre-training principle, it would be useful to sequence the instruction to begin with descriptions of the key concepts or elements before describing how they interact. According to the modality principle, it would be useful to use narrated animation rather than animation with on-screen text. Furthermore, the designer should take learner characteristics into account and be aware that well-designed multimedia instruction may be most effective for low-experience learners.   

    What Are the Limitations of Current Research and What Are Some Productive Directions for Future Research?

T he principles described in this chapter are subject to limitations inherent in the nature of the task and in the dependent measures. Concerning the nature of the task, most of the research studies involved short narrated animations presented in a controlled laboratory environment with an immediate test. Research is needed in which the principles are tested within more ecologically valid environments, such as with students in their classrooms, as was done by Harskamp et al. (2007 ). Concerning the dependent measures, we have focused on measures of transfer because we are most interested in improving learners’ understanding. However, a central assumption underlying the principles described in this chapter is that they work because they effectively reduce cognitive load. Thus, it would be useful to include direct measures of cognitive load in future research (Brunken, Plass, & Leutner, 2003;Brunken, Seufert, & Paas,   2 010;DeLeeuw & Mayer, 2 008) .Further research work is also needed to better understand the conditions under which each of the principles is most effective. Concerning the segmenting principle, some of the studies focused solely on allowing the learner to control the pace of presentation, whereas others focused on allowing the learner to control both the pace and order of presentation. Research is needed to determine the relative effects of learner control of presentation pace and learner control of presentation order, and whether such effects depend on the characteristics of the learner and on the characteristics of the learning task. For example, inexperienced learners may lack the metacognitive skills to make effective decisions about the order of presentation, whereas experienced learners may be able to make such decisions effectively. The role of the learner’s prior knowledge is examined in detail by Kalyuga in Chapter 24 . Also concerning segmenting, research is needed to determine the most effective size of a segment and whether optimal segment size depends on the characteristics of the learner and the learning task. In some of the studies reported in this chapter, the segments were fairly short – approximately 10 seconds of animation along with a few sentences of narration – and the segments described meaningful steps in a process. Having segments that are too small may distract and irritate some learners, whereas having segments that are too long may result in cognitive overload.Concerning pre-training, research is needed to extend the principle beyond cause-and-effect systems. In the studies reported in this chapter, the learning goal was to understand how some mechanical or physical system works. In such situations, it was useful to know the names and behaviors of each component before learning about the causal model. Further research is needed to determine whether similar benei ts of pre-training occur for different kinds of lessons, such as an explanation of how to solve a mathematics problem or an analysis of a historical controversy. Concerning modality, there may be situations in which on-screen text would be helpful, such as for lessons with complex technical terms or long segments of unfamiliar text or for learners who are non-native speakers or who have hearing disabilities. Further research is needed to determine the conditions for the modality principle and how they relate to cognitive theory.   

DMU Timestamp: November 12, 2020 20:50





Image
0 comments, 0 areas
add area
add comment
change display
Video
add comment

Quickstart: Commenting and Sharing

How to Comment
  • Click icons on the left to see existing comments.
  • Desktop/Laptop: double-click any text, highlight a section of an image, or add a comment while a video is playing to start a new conversation.
    Tablet/Phone: single click then click on the "Start One" link (look right or below).
  • Click "Reply" on a comment to join the conversation.
How to Share Documents
  1. "Upload" a new document.
  2. "Invite" others to it.

Logging in, please wait... Blue_on_grey_spinner